Big Data Analytics Paving Way For Predictive Consumer Targeting


The rising importance of Big Data has coincided with the growth of data.  According to research firm IDC, all data has grown at a rate of 87% from 10 EB (1 EB = 10^18) in 2009 to 80EB in 2014. However, on breaking this data into its components - structured data and unstructured data – the results were found to be even more astonishing. The break up reveals that while structured data has remained more or less stagnant, unstructured data has grown by almost 90% in the same period of time.

Unstructured data, which can be anything from email, social, video to text, is invaluable as it provides deep insights into customer behavior helping marketers avoid falling into the trap of stereotyping their customers. However, the problem is in making sense of this data. Computers are good only if the data is formatted or “structured” into neatly arranged “rows and columns”. SQL (S for Structured; Q for Query; L for Language) can parse through databases with millions of rows and columns and return the desired result in seconds. This is where Big data steps in. It uses tools like predictive analysis, in-memory computing and advanced heuristics, providing a more complete view of the customer, revealing customer intent and showing the predictive path to future behavior.

Today, the customer has information; not only about the product but also about the alternatives available in the market. As such, the customer will “switch” if their expectations are not met. In this fast emerging customer environment, it is not enough to meet customer expectation; businesses must strive at all times to “exceed” customer expectations to stay relevant in the dynamic markets of today. A good example of how big data and predictive analysis is changing customer experience is a 156 year old iconic luxury brand, Burberry.  By investing in big data Burberry is enabling its staff to exceed customer expectations. By connecting the dots between social and customer’s instore buys, Burberry is able to identify its customer the moment they walk into the store.

Marketers have understood that the difference between customer stereotyping and customer segmentation is data. This means that data empowers them to move beyond thinking “if a consumer belongs to this group, she will behave this way”, opening up their minds to new and innovative ways to influence consumers.Having started off with demographic segmentation, marketers have moved along psycho-demographic segmentation, attitudinal segmentation and have reached behavioral targeting to narrow down to a campaign segment.  Behavioral targeting makes use of past expression of interest in a particular product, in the form of a purchase, website visit or keyword search to identify future targets of the product’s marketing campaigns. The central flaw in behavioral targeting lies in the fact that it is reactionary. There are no proactive measures to identify consumers who have not fallen into any of the behavioral categories identified by the marketer. This means that prospective buyers who have not clicked on an ad, purchased the product in the past or visited the product site are completely overlooked while considering the target group.

Predictive targeting has come to the limelight to clear up these shortcomings and subsequently announce the comparatively early demise of behavioral targeting. Evolution of machine learning algorithms have removed barriers in considering a vast range of different attributes in predicting the level of interest each consumer has in a product. For example, eBay with more than 100 million customers is often described as a self-regulating economy with transactions running into hundreds of billions of dollars. Like any free market economy, eBay uses listing policies, price cuts to stimulate consumption (or in this case buying behavior), but with so many transactions happening in real time it needed a powerful system for crunching data. Predictive analysis helps in distinguishing the signal from the noise by tracking thousands of business metrics, forecasting for tomorrow and sending an alert if there is any deviation.

 Thus the predictive algorithm, not the marketer, identifies and continually updates the segment based on each consumer’s likelihood to convert their interest into a purchase. This saves marketers from committing the error of ignoring the innumerable people who have not explicitly expressed interest in the product. The approach analyzes the attributes of a sample set of people and overlays the insights onto the entire universe of consumers, online and in real-time. The key win of the method lies in the fact that it combines all previous approaches of targeting be it psycho-demographics, attitudinal or behavioral – as all this data are captured in the attributes that are used to profile customers.  It also takes into account consumer behavior across obscure and low-key attributes, previously considered as noise. In all ways, predictive targeting literally leaves no stone unturned.

Yet another prominent virtue of predictive targeting is that it handles the ever-evolving consumer profile data quite well. Consumer behavior is highly dynamic and a highly evolved and flexible analytic algorithm is the only way for marketers to get their head around this complexity. This massive data that may seem overbearing to marketers is converted to powerful insights by analytical algorithms, for building highly cohesive customer profiles. The campaigns made using these profiles can also hence be made eminently evolving, with adjustments made to them to accommodate the changing consumer trends. Thus segment composition becomes dynamic, according to the changing behavior of the consumers as well as the progressive strategies and objectives of the marketers.

A “predictive” approach replaces the “gut feel” with a “data” driven approach where outcomes are known or at least understood. The following are some of the use case scenarios for predictive analysis:

· Churn Management: Identifies customers who are on the verge of leaving. Since new customer acquisitions are very high, churn management can help businesses retain customers by special offers. Caesars uses big data analytics to identify regular customers who are on a losing streak, providing them with a free pass to spa or restaurant nullifying churn.  AmEx uses sophisticated predictive models to identify 24% of the accounts that would close within the next 4 months 

· New customer acquisition: Identifying prospects providing the most lifetime value to the company. Express Scripts, which processes pharmaceutical claims, found that those who needed medicines also tended to forget their medication so they designed beeping medicine bottle caps and mobile reminders.

· Customer contact: Predicting the best deal, the best time and the best channel after studying customer behavior as well as social interactions. The Commonwealth Bank of Australia (CBA) uses big data and analytics to provide customized search experience on its website. Website visitors are provided with more relevant offers related to what they have been searching for on the website. For example, a user searching for properties are provided information related to loans and insurance.

· Managing maintenance cycle: Helps to predict outlays needed for maintenance and thus preventing downtime. Tesco PLC, the supermarket chain, keeps a tab on 70 million refrigerator related data points allowing it to schedule timely maintenance 

· Sentiment analysis:Using analytics to get a bird eye’s view of your company helping in managing the reputation of your company. When a customer tweeted Morton’s the Steakhouse for a dinner to be sent to the Newark airport, the steak house pulled data of what he ordered previously, figured out what flight he was on, and then sent a delivery person with the order.   

· Upselling: predictive analysis can be used to upsell to your customers helping you to increase the revenue of your company as well as increase value to your customers.  

If there’s one thing that marketers agree on despite their conflicting methodologies, it is that there is nothing called too many new customers. Predictive targeting is equally instrumental in getting new customers as well as converting leads into paying customers. The use of artificial intelligence and machine learning to optimize audience selection and to enable accurate targeting of the right prospects at the right time to maximize impressions and conversions has indeed come as a boon to the marketing field. Predictive marketing tools will thus surely take away an important share of technology budgets that CMOs plan to deploy in the year of 2016.